Texture Featuring and Indexing for Region-Based Image Retrieval
Speaker: Ying Liu, ITEE
When: 2003-06-30 13:00:00
Venue: 78-622
Host: Xiaofang Zhou
Abstract:With a rapid increase in the size of digital image collections in
all walks of life, an efficient image retrieval technique is highly
on demand. This has led to content-based image retrieval (CBIR)
techniques - retrieving images on the basis of automatically-derived
features such as colour, texture and shape. As users accessing CBIR
systems are often interested in regions, region-based image
retrieval (RBIR) methods are developed to represent images at
region-level, which is intended to be close to the perception of
human visual system.
Among contents based features, texture is fundamental for image
classification. Our study focuses on texture features.
Most current RBIR approaches use small database and do not consider
indexing. As image database become larger and larger, efficient
indexing of high-dimensional image features is required for fast and
efficient retrieval. Existing indexing algorithms are designed
without considering the properties of image data and the
requirements of RBIR systems. We are going to design an efficient
RBIR system based on effective image featuring and indexing. We
expect that compared with current RBIR approaches which do not
considering indexing, our system will provide faster retrieval with
little or no degradation in retrieval quality. On the other hand,
compared with indexing algorithms designed without considering the
properties of image data, our system intends to provide more
efficient retrieval specially for image data.
Biography:(biography unavailable)
Type: Ph.D confirmation
Contact:Xiaofang Zhou, seminar host (zxf@itee.uq.edu.au)
or Guido Governatori (ITEE seminar co-ordinator)
(guido@itee.uq.edu.au)